Michael Niklas

A sidenote: - I have removed the `_netCDF4` module path from the reprs, I doubt that should affect anyone but one could consider this a breaking change. - I was...

> @headtr1ck please see my questions in the review I don't see a review? Are you sure you published it?

Don't know if anyone noticed, but `dtype_is_complex` seems to be wrong? It only works if you pass string literals like `"c8"` or `"c16"`, but not if you pass a numpy...

@jswhit any reason, `__has_parallel_support__ ` and `__has_ncfilter__` were not imported at `netCDF4` level (only available in `netCDF4._netCDF4`? I imported them now, but can undo that.

> Ugh, I wish I'd found this PR sooner. I've been working on the same thing but as a [stubs package](https://github.com/RandallPittmanOrSt/netcdf4-python-stubs). Looks like I have some things to learn from...

I think having the docstrings in two places sounds like a bad idea. This would require some CI checks to test if they are the same etc and would make...

I would say, we focus on typing first in this PR and then open a separate issue for the docstrings. Maybe there is a way to copy over the docstrings...

I would say I leave it like this. The Variable proposal with all the Descriptors looks quite hacky and is beyond my knowledge. I propose to merge as is, and...

> [stubtest](https://github.com/Unidata/netcdf4-python/actions/runs/9686631247/job/26732005277?pr=1302) is failing Can you try again please?

> This means we could replace `Union[]` with `|` everywhere in the type hints right? Shouldn't pyupgrade / ruff do that for us?